screencast
Pass
Audited by Gen Agent Trust Hub on Sep 6, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external data sources such as footage metadata, project files, and production documents as part of its intake process. This creates a surface where instructions embedded in these external files could potentially influence the agent's logic.
- Ingestion points: Evaluates metadata, project files, and user-supplied production documents (SKILL.md).
- Boundary markers: No explicit use of delimiters or instructions to ignore embedded prompts in external data is defined.
- Capability inventory: Instructions allow for direct execution of media and platform actions if supported by the environment's tools (SKILL.md).
- Sanitization: The skill does not specify validation or filtering steps for the ingested media metadata or document content.
Audit Metadata